Publications
2022
-
Representing and Analyzing Sequential Satellite Mission Design Decisions through Anisomorphic Trees and Directed GraphsAIAA SCITECH 2022 Forum, Jan 2022In recent years, decision trees have been an uncommon approach to design space exploration, and meta-heuristics and machine learning have been the primary approach. Although these approaches have proven effective, decision trees have the unique property that allows us to explicitly traverse the design space sequentially, in a human-understandable form. The problem is the order of the decisions in the tree is not pre-defined but chosen by the user, leading to anisomorphic trees and inconsistent performance during design space exploration. In the past, we viewed this as a drawback of decision trees that designers needed to overcome. However, in this paper, we study anisomorphic decision trees representing a satellite mission design problem to glean insights into design decision-making in general. In this case study, we look at an earth observation satellite mission design formulated as an instrument-to-orbit assignment problem and quantify the effects of design decision tree characteristics. This work has applications in design space exploration and design automation when working on complex design problems by improving our understanding of how to represent and search the space. Additionally, we define the relationship between all anisomorphic trees for a problem and a novel structure that we call the Design Space Directed Graph, contributing to the general understanding of design decisions and their mathematical representations.
